While the fix results in consistent cell size proportions, the effective offsets for users which previously set the offset and glyph_offset options to target a display with a scale_factor > 1.0 will change. Is this acceptable?

The change is a bit questionable, because some may desire to adjust based on physical pixels and not be constrained by logical pixel grid. It could also weirdly work with non integer scale factors.

I agree that the simplicity of adjusting based on physical pixels is great and should be kept. I still think that there is some value in having an option which allows to use offsets and move a window between displays without changing the window's content (I've attached a screenshot with the same window moved between two displays with different scale factors). Would it be OK to add an option for logical pixels and keep the default based on physical pixels?

Regarding non-integer scale factors: The current fix truncates the calculated physical offsets towards zero but this could also be changed to rounding instead.

I agree that the simplicity of adjusting based on physical pixels is great and should be kept. I still think that there is some value in having an option which allows to use offsets and move a window between displays without changing the window's content (I've attached a screenshot with the same window moved between two displays with different scale factors). Would it be OK to add an option for logical pixels and keep the default based on physical pixels?

Maybe due to us scaling the font sizes, actually scaling the offsets is more desirable, but it could lead to situation where you can't offset the way you want. I guess we also have something similar when it comes to padding, and maybe some other metrics, since those are generally using the physical pixels.

Yes, but if that's important there could be a config option to offset based on logical vs on physical pixels. What's your suggested way of proceeding with this, @kchibisov?